Home | About Us | Contacts | Help | Feedback | SEARCH Global | Australia | NSW | Vic. | Qld | WA | SA | Tas. | ACT | NT | Ant. | Weather & Warnings | Hydrology | Climate | Numerical Prediction | About Services | Learn About Meteorology | Registered User Services |

El Niņo - Southern Oscillation: Aftermath of the 1997/98 event and prediction of future events  

Neville Nicholls

Bureau of Meteorology Research Centre, Melbourne, Australia

Scientific lecture presented at the 13th Congress of the World Meteorological Society, Geneva, 21 May 1999

The 1997/98 El Niņo was one of the most severe El Niņo events of the past century. It was also the best observed El Niņo event, received the most extensive and closest media interest and public attention, had more impacts than ever before blamed on it, and had many more people and organisations than ever before predicting and commenting on its likely progress. We are all El Niņo experts now! But what did we learn from the event, and will we use these lessons to improve our reactions and forecasts for the next El Niņo? Trenberth (1998) and McPhaden (1999) describe the development of the 1997/98 El Niņo - Southern Oscillation, and the scientific issues that arose from the event. Here the concentration is on the forecasting of the event and, especially, how to improve the use of climate forecasts based on the El Niņo - Southern Oscillation. 

The Australian Bureau of Meteorology's National Climate Centre has been issuing seasonal climate forecasts of rainfall, based on the El Niņo - Southern Oscillation, for more than a decade. Despite this experience, and despite recognition early in 1997 that an El Niņo was developing and likely to be severe, we learned much from the 1997/98 El Niņo. The event demonstrated, inter alia, that the forecast task requires more than just the development of a skilful forecast model – a clearer focus is needed on the dissemination and use of climate forecast information. 

What can we say about the 1997/98 El Niņo? First, by any measure, it was severe, and many serious impacts were associated with it. This event was the first in which many national and international organisations, (eg, FAO) reacted during the course of the event. Such rapid response seems likely to be a feature of future El Niņo events. The global impacts of the 1997/98 El Niņo have been estimated by NOAA (see Table 1). The event is estimated to have caused losses of around $US35,000 million. As well, the human health impacts were large - for instance the haze from forest fires in southeast Asia affected the health of many millions. Improved forecasts and optimal use of our current knowledge could reduce these impacts. In some cases, during 1997/98, it seems our understanding of El Niņo did mitigate its impacts.

The timely media coverage of the El Niņo impacts was one feature that distinguished the 1997/98 El Niņo from previous events. Although there was considerable media coverage of the 1982/83 event, much of this was published at the end of the event. The 1997/98 coverage, as well as being in "real time", also was more thorough, more accurate, and, indeed, more useful. Excellent articles describing the mechanics of the El Niņo - Southern Oscillation appeared in many news magazines. The science of El Niņo, and media and public interest in it, has developed much in recent years. 

At times the concerns raised by media coverage of the El Niņo were perhaps more extreme than the event warranted. The tone of media articles contributed to a feeling of panic in rural Australia, a feeling not justified by realistic assessments of the relationship between El Niņo and drought, nor indeed by our forecasts. One lesson we learnt during 1997 is that we need to improve our interaction with the media and users, to ensure the coverage is balanced (Nicholls and Kestin, 1998). Table 2 lists the terms we used during 1997/98, to try to convey to the media our conviction that much of Australia was more likely than usual to experience drier than normal conditions. The variety of terms used to describe both the event (rainfall in the lowest tercile) and the probability of that event happening, and the lack of formal definitions of these descriptors, complicated the dissemination and use of forecasts – we need a formal definition of terms such as "likely", for the purposes of forecast formulation and dissemination. These definitions must match how these terms are interpreted by the general public and media, if we are to convey our message accurately to the public and decision makers. One way of improving understanding would be to always include numeric values of the probabilities after a verbal description, eg.: "There is a significant chance (70% or more) of below median rainfall". 

The problem with the forecast terminology was exaggerated by the presence of many forecasts from various sources, all using loosely-defined, but different, terminology. This led to some users ignoring all information regarding the El Niņo, because they felt the "scientists couldn't agree". The establishment of formal definitions of probability descriptors in climate prediction could be done by CLIPS – a WMO lead role in this area would certainly reduce confusion caused by multiple forecasts using a variety of terminology. WMO has a natural role to play here, since many forecasts are sourced from outside the "target" country. 

Much of the increased media coverage reflected our growing confidence, since the 1982/83 event, that many climate impacts were attributable to the El Niņo, and to our growing confidence that we understood and could predict the El Niņo, its climate effects, and its impacts. However, the recent El Niņo perhaps has left an inflated impression of our prediction ability. An article in the journal Science asserted that "Predictions of the most recent El Niņo were widely regarded as a stunning success" and that "if meteorologists had dared to rely more heavily on their computer models, those predictions could have been even better". In fact, few models predicted an El Niņo until the event had started in April 1997; some models refused to predict an El Niņo even when the event was at its peak; no model predicted that it would be anything more than a moderate event; and perhaps the best overall predictions came from a very simple statistical scheme rather than a complex coupled ocean-atmosphere model (Barnston et al., 1999). Once the event was under way the models, in general, predicted its subsequent evolution, although they were slow to predict the demise of the event. 

Once the event was established, by mid-1997, the climate impacts of the event were predictable, in general, by simple statistical methods. The pattern of climate anomalies, generally dry in the west Pacific, wet in the east, was as expected. Simply knowing that an El Niņo had commenced allowed many potentially useful predictions of climate variations. How good were the models in predicting climate, once the El Niņo was under way? They also exhibited skill, in some regions. However, as was the case with the models designed to forecast the El Niņo itself, there is still no strong evidence that models, as yet, consistently and substantially outperform statistical methods. 

Perhaps we are at the point that weather prediction was at 20-30 years ago, when models were starting to match the accuracy of subjective or statistical forecasts. If so, then perhaps in the next few years climate models may start to outperform statistical schemes. Certainly there are theoretical reasons for believing that climate models should be able to overcome some of the limitations of statistical models. The models already do show some skill, and the 1997/98 El Niņo will be remembered as the first in which models were widely used to support operational climate prediction. For the immediate future, however, much of the evidence that an El Niņo has commenced will still come from observations of the Southern Oscillation Index (SOI) or of east Pacific sea surface temperatures. The monitoring of these temperatures has improved dramatically since the 1982 El Niņo, largely due to the Tropical Atmosphere-Ocean buoy array (McPhaden, 1999). WMO, through programs such as TOGA and more recently CLIVAR and GCOS, has been instrumental in improving the way we monitor El Niņo. The El Niņo - Southern Oscillation is one of the principal research areas of the CLIVAR program, and this should lead to improved models and forecasts. 

A realistic assessment of our current ability to predict El Niņo and its climate effects is necessary because of the way we, and our public, deal with uncertainty. There is considerable evidence in the psychological literature that people do not handle uncertainty and probabilities at all well. A group of psychological factors called cognitive biases confound our attempts to communicate and understand uncertainties including the uncertainties associated with climate (Nicholls, 1999). Since we must deliver climate forecasts in terms of probabilities, because of the chaotic nature of the climate, we must also learn how people interpret, and misinterpret, these probabilities. A selection of these biases are described in Table 3. These cognitive biases affected the way El Niņo forecasts were received and interpreted in Australia during 1997. For instance, two biases are "availability" and "anchoring". During 1997 many press articles on the El Niņo described the severe impacts of the 1982/83 event. Users then had great difficulty adjusting their expectations of the impacts of the 1997 El Niņo away from what they had experienced during 1982. This occurred even when they were reminded that the 1982 impacts were very extreme, compared with historical impacts of the El Niņo on Australia. The "availability" of the reports about the 1982/83 impacts led users to "anchor" to the 1982 impacts, and they subsequently could not "adjust" away from that anchor sufficiently. Nicholls (1999) discusses ways to avoid these cognitive biases – the "anchoring" problem, for instance, could be reduced by ensuring that a variety of El Niņo events, with varying degrees of impact, are discussed in the context of a forecast based on El Niņo. Other ways to reduce the likely effect of cognitive biases are listed in Table 3. We need to work on, not just improving our monitoring and predicting of the next El Niņo, but on innovative ways to present these predictions to the public, if we are to realise the benefits of the last few decades of scientific research. 

I want to stress the impact of this research. We have come a long way very quickly and I think this can be illustrated by noting Australian reactions to the last four or five El Niņo events. In 1972 we did not even realise we were in an El Niņo until 1973! We were six months into the 1982/83 event before forecasts were issued, and it was then too late to avoid much of the damage. In the moderate 1991 and 1994 events we did issue forecasts not long after the start of the events, and in 1997 we did better than ever before. This improvement in timeliness reflects improved monitoring, improved telecommunications, improved statistical efforts to related El Niņo to its climate effects and improved modelling and scientific understanding of the phenomenon. Two articles on El Niņo events, published in the National Geographic, remind us of how far we have come, and how quickly. A 1984 article on the 1982/83 event catalogues the destructiveness of the El Niņo, whereas a 1999 article on the recent event focuses more on our improved ability to monitor and predict the phenomenon. This is a remarkable advance in only 15 years, in a field as complex as climate. We have a considerable way yet to go, but we should not ignore the very real and very important progress that has been made in the past few decades. 

We need, obviously, to improve our monitoring and predictive capability. But equally important, we need to improve our communication and liaison, so users can take full advantage of what we already know and can predict about El Niņo. This will require better partnerships with the media to ensure the message is told accurately and intelligibly, work with psychologists to ensure that the underlying uncertainties are fully appreciated and accounted for by decision makers, and work with impact communities such as farmers to ensure the real impact of El Niņo is understood and acted on, without panic but also without underestimating the risks. We need clearer ways of expressing the expected accuracy and risks of the forecasts. A major task for NMHSs will be to "filter" the numerous messages and forecasts regarding El Niņo – there will be even more of these in the next El Niņo than was the case in 1997/98, and these could lead to increased confusion on the part of the public and interested groups. 

Although El Niņo is important, its opposite extreme, La Niņa, can also be destructive and deserves attention. And there are other climate phenomena we can monitor and may be able to predict. So there is still much to be done, before we can confidently assert that we can predict interannual climate variations. But let us not undervalue what has been achieved already and has been demonstrated in the 1997/98 El Niņo. Michael Glantz, from NCAR, has predicted that forecasting El Niņo will be "science's gift to the 21^st century". The 1997/98 event confirmed to us, and demonstrated to our public, that we can use the El Niņo in forecasting, just in time for the start of the new century. Michael Hall, from NOAA Office of Global Programs, has pointed out that the application of El Niņo forecasts will be the biggest advance in the use of climate information since humans applied the annual cycle to agriculture. Again, the use of forecast information during 1997 showed we can do this – apply what we already know about El Niņo to mitigate climate damage and take advantage of climate opportunities. The challenge for us is to do it even better next time. In meeting this challenge, WMO will be the crucial organisation, and the success of WMO's climate programs will determine our success.

References 

Barnston, A. G., M. H. Glantz, and Y. He, 1999. Predictive skill of statistical and dynamical climate models in SST forecasts during the 1997-98 El Niņo episode and the 1998 La Niņa onset. Bulletin of the American Meteorological Society, 80, 217-243. 

Janis, I. L., 1982. Groupthink: Psychological studies of policy decisions and fiascos. Houghton-Mifflin, 379 pp. 

McPhaden, M. J., 1999. Genesis and evolution of the 1997-98 El Niņo. Science, 283, 950-954. 

Nicholls, N., and T. Kestin, 1998. Communicating climate. Climatic Change, 40, 417-420. 

Nicholls, N., 1999. Cognitive illusions, heuristics, and climate prediction. Bulletin of the American Meteorological Society, 80, 1385-1398. 

NOAA-OGP, 1999. An experiment in the application of climate forecasts: NOAA-OGP activities related to the 1997-98 El Niņo event, U.S. Dept. of Commerce, January 1999, 134 pp. 

Trenberth, K. E., 1998. Development and forecasts of the 1997/98 El Niņo: CLIVAR scientific issues. CLIVAR Exchanges, 3, 4-14.

Table 1. Global impacts of the 1997/98 El Niņo - Southern Oscillation event (from NOAA-OGP, 1999).

Table 2. Terms used in Australian Bureau of Meteorology Seasonal Climate Outlooks during 1997/98 El Niņo (Tahl Kestin, pers. comm.). 

Table 3. Some "cognitive biases" likely to affect the preparation and use of climate forecasts, and ways to reduce their effect (from Nicholls, 1999).